Call Duration Estimation for Dynamic Queue Prioritization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing telephone support systems face inefficiencies in call queue management due to calls being handled in the order of arrival, leading to prolonged wait times for customers when service agents are overwhelmed.
Innovation Solution
Implementing a method that utilizes a Hidden Markov Model (HMM) and Gradient Boosted Machine (GBM) regression model to estimate call durations based on historical transaction, account, and call history information, allowing for priority assignment and optimized call handling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If calls are handled in the order of arrival (FIFO), then the system maintains simple queue management, but customer wait times are prolonged and efficiency decreases
Solution Approach 1:
The system performs preliminary analysis of incoming calls by examining historical information, transaction data, and call patterns before the call is actually handled. This advance preparation enables the system to predict call duration and assign priorities before agents need to service the calls, allowing optimized queue positioning rather than simple FIFO ordering.
Solution Approach 2:
The call queue management transitions from a static FIFO approach to a dynamic priority-based system. Call priorities are not fixed but are determined by real-time analysis of multiple factors including historical behavior, transaction context, and predicted duration. This dynamic reordering optimizes both agent productivity and customer wait times by strategically positioning calls in the queue.
2Productivity
If the system assigns priorities based on estimated call durations, then call handling efficiency improves, but system complexity increases due to multiple models and data processing
Solution Approach 1:
The complex prediction task is segmented into distinct functional components: a Hidden Markov Model segment that processes historical and transactional data to identify patterns, and a regression model segment that translates these patterns into call duration predictions. This segmentation allows each component to specialize in specific aspects of the prediction problem, making the overall complex system more manageable and maintainable.
Solution Approach 2:
The Hidden Markov Model acts as an intermediary between raw historical data and the final duration prediction. It processes complex sequential patterns in customer behavior and transforms them into a format that the regression model can effectively utilize. This intermediary layer simplifies the overall system architecture by breaking down the complex prediction task into manageable stages with clear interfaces.
Data Source
AI summary
A method for call processing includes: receiving a first incoming telephone call from a first person; retrieving historical information that relates to the first person; forecasting, based on the historical information, a next customer state that relates to the first person; and estimating, based on the forecasted next customer state, a potential duration of the first incoming call. The estimated call duration is then usable for assigning a priority to the first incoming call and determining an order of handling the first incoming call with respect to each of a plurality of other incoming calls based on the assigned priority.